bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

1 Rare morph benefit from reduced aggression from con- and hetero-specifics

2 Running title: Reduced aggression benefits rare morph mbuna

3

4 Alexandra M. Tyers*, Gavan M. Cooke & George F. Turner

5 School of Biological Sciences, Bangor University, Deniol Road, Bangor. Gwynedd. Wales. UK. LL57 2UW

6 * Current address: Max Planck Institute for Biology of Ageing, Joseph-Stelzmann-Straße 9B, 50931, Köln

7

8 Corresponding author: A.M. Tyers, [email protected]

9

10 Abstract

11 Balancing selection is important for the maintenance of polymorphism as it can prevent either fixation of one

12 morph through directional selection or genetic drift, or speciation by disruptive selection. Polychromatism can

13 be maintained if the fitness of alternative morphs depends on the relative frequency in a population. In

14 aggressive , negative frequency-dependent antagonism can prevent an increase in the frequency of rare

15 morphs as they would only benefit from increased fitness while they are rare. Heterospecific aggression is

16 common in nature and has the potential to contribute to rare morph advantage. Here we carry out field

17 observations and laboratory aggression experiments with mbuna cichlids from Lake Malawi, to investigate the

18 role of con- and heterospecific aggression in the maintenance of polychromatism and identify benefits to rare

19 mores which are likely to result from reduced aggression. Within species we found that males and females bias

20 aggression towards their own morph, adding to the evidence that inherent own-morph aggression biases can

21 contribute to balancing selection. Over-representation of rare morph territory owners may be influenced by

22 two factors; higher tolerance of different morph individuals as neighbours, and ability of rare morphs to spend

23 more time feeding. Reduced aggression to rare morph individuals by heterospecifics may also contribute to rare

24 morph advantage.

25

26 Key words: Malawi, , blotch polymorphism, aggression, rare morph advantage

27

28

1 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

29 Introduction

30 Permanent polymorphism, the presence of multiple genetically determined morphological or behavioural

31 phenotypes within a population, is common in nature and indicates some type of selective balance between

32 morphs. Balancing selection is important for the maintenance of polymorphism as it can prevent either fixation

33 of one morph through directional selection or genetic drift, or speciation by disruptive selection (Huxley 1955;

34 Wellenreuther et al. 2014; Kim et al. 2019). Polychromatism (colour polymorphism) can be maintained if the

35 fitness of alternative morphs differs in time or space in heterogeneous environments, or if the fitness of a

36 phenotype depends on its relative frequency in a population (Hughes et al. 2013; Pérez i de Lanuza et al. 2017;

37 Surmacki et al. 2013; Svensson 2017; Henze et al. 2018).

38 In many taxa, species-recognition cues have diverged through reproductive or antagonistic character

39 displacement to reduce hybridisation or unnecessary exertion and risk of injury among heterospecifics which

40 are not in direct competition for mates or resources (Seehausen & Schluter 2004; Grether et al. 2009). Rare

41 colour morphs can benefit from lack of recognition by receiving less mating-related harassment (Takahashi et al.

42 2010) or less intrasexual aggression from conspecifics (Dijkstra et al. 2008; Lehtonen 2014; Pérez i de Lanuza G

43 et al. 2017; Scali et al. 2020). In aggressive species, negative frequency-dependent antagonism, generated

44 through either evolution of an own-morph bias (Dijkstra et al. 2008; Lehtonen 2014; Scali et al. 2020) or by a

45 dynamic common morph bias based on experience (Bolnick et al. 2016), can prevent an increase in the

46 frequency of rare morphs as they would only benefit from increased fitness (due to reduced aggression) while

47 they are rare (Seehausen & Schluter 2004; Dijstra et al. 2007; Bolnick et al. 2016).

48 The existence of conspecific aggression biases does not preclude heterospecific aggression completely.

49 Indeed, resent studies suggest that heterospecific aggression as a result of resource competition and

50 reproductive interference may be more common than previously assumed (Grether et al. 2009; Drury et al.

51 2020). Regardless of whether heterospecific aggression is due to convergence in territorial signals among

52 species competing for resources or due to misdirection of aggression because closely related species still share

53 similar signals (Losin et al. 2016), in a variety of taxa aggression is often higher among more similar coloured

54 than more differently coloured species (Genner et al. 1999; Pauers et al. 2008; Anderson & Grether 2010; Losin

55 et al. 2016). In taxa where multiple ecologically and phenotypically similar species co-exist in the same habitat

56 there is therefore potential for rare morphs to benefit not only from reduced conspecific aggression, but also

2 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

57 from reduced heterospecific aggression. A recent study of Midas cichlids, however, demonstrated increased

58 aggression towards rare heterospecific morphs and suggested that this disadvantage may help to explain their

59 lower frequency in natural populations (Lehtonen et al. 2015). The role of heterospecific aggression in relation

60 to polychromatism requires further exploration to improve our understanding of how this may contribute to its

61 evolution and maintenance.

62 The mbuna cichlids of Lake Malawi (and the closely-related ecologically-similar Mbipi of Lake Victoria)

63 provide an excellent system for the investigation of colour polymorphism . Mbuna inhabit densely packed multi-

64 species communities in the shallow-waters and identify conspecific mates and rivals predominantly by their

65 species-specific colour and pattern (e.g. Seehausen & van Alphen 1998; Couldridge & Alexander 2002; Jordan

66 2008; Pauers et al. 2008). Several species display a polychromatism characterised by the presence of rare

67 “blotched” morph individuals, which occur at different frequencies in different species and populations (Lande

68 et al. 2001; Ribbink et al. 1983; Konings 2007). While it is likely that predation has played some role in the

69 evolution of this polychromatism (Seehausen et al. 1999; Streelman et al. 2003; Maan et al. 2008), and mate

70 choice may have been involved in the evolution of (partial) sex-linkage (Seehausen et al. 1999; Lande et al.

71 2001; Roberts et al. 2009), it is thought that intrasexual competition plays a large role in its maintenance

72 (Dijkstra et al. 2008; Dijkstra et al. 2009b). Although in most species the frequency of rare morphs remains

73 relatively low in all populations, in some, for example callainos at Thumbi West Island in Lake

74 Malawi, rare morphs can occur with higher frequency, which allows greater ease of observation and collection.

75 Here we used this population to conduct field observations and laboratory behavioural experiments to test

76 alternative hypotheses regarding aggression biases: Do both morphs preferentially direct aggression towards

77 the common (presumably ancestral) morph, or is there an own-morph bias? An own-morph bias could be

78 sufficient to maintain polymorphism through negative frequency-dependent selection, while a common-morph

79 bias would suggest that an additional frequency-dependent process would be necessary to limit an increase in

80 the number of rare morph individuals. We also test for aggression biases towards the common and rare morph

81 from a closely related heterospecific to assess whether this may contribute to balancing selection. We aim to

82 identify potential benefits to rare morphs, which may occur as a result of receiving less aggression, in the

83 natural environment. Additionally, as differences in selection pressures on each sex, due to differences in the

84 type of competition they experience (competition for mates among males and competition for non-mating

3 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

85 resources among females) can result in sex differences in the types of aggressive behaviour used during

86 contests (Arnott & Elwood 2009), we also test for sex-differences in aggressive behaviour and aggression biases.

87

88 Methods

89

90 Study system

91 (= callainos or Metriaclima callainos) is a member of the ‘mbuna’ complex

92 of rocky shore cichlid fishes endemic to Lake Malawi. Populations of M. callainos are found in shallow water,

93 with peak population density between 3 – 10m (full range 0 – 25m). Their natural range is confined to the

94 northern end of Lake Malawi, where they are often found in sympatry with the more widely distributed

95 ecologically similar congeneric M. zebra. However, due to human mediated translocations, they are also found

96 in some southern areas. Phenotypes of common and rare morph mbuna differ between species, but within

97 populations, the common morph is often BB (black vertical melanin bars on a blue/dark background) or solid

98 blue/dark body colour, while rarer morphs have a disrupted melanin pattern of many or few blotches/spots on

99 a light (orange/pink/white) body. Blotch polychromatism is not present in all M. callainos and M. zebra

100 populations; at some localities only the plain blue (B) and BB morph are found, whereas at others, these

101 common morphs may occur along side rare white (W) and orange-blotch (OB) and very rare white-blotch (WB)

102 and orange (O) morphs. In this study we focus on a well established translocated population of M. callainos at

103 Thumbi West Island in the Lake Malawi National Park in the southwest arm of the lake which has both B and W

104 morphs. The likely source population of the M. callainos at Thumbi West is Nkhata Bay, where they co-occur

105 with a population of M. zebra comprised of BB, OB and O morph individuals (fig. 1).

106 All fish were wild caught: M. callainos and M. zebra from Thumbi West Island (TW) in July 2010, M. zebra

107 from Nkhata Bay (NB) and Chiofu Bay (CB - naïve to M. callainos in the wild and lab) in 2009. Males and females

108 were used in this study, partially because of the lower number of rare males, but also because both male and

109 female aggression biases have previously been suggested to be important in colour polymorphism maintenance

110 in cichlids (e.g. Dijkstra et al. 2008). Furthermore, unlike many species with blotch polychromatism, this one is

111 less strongly female limited, as numerous white M. callainos males were found at the study/collection site.

112

4 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

113 Figure 1. a) Mbuna community at Thumbi West Island. Territorial male (bottom) and female (top) of b) M.

114 callainos Blue morph, c) M. callainos White morph, d) M. zebra BB morph. Coloured squares correspond to

115 colours used for the different morphs in results plots.

116

117 Field observations (excluding aggression)

118 Frequency of blue and white morph M. callainos at Thumbi West Island

119 Snorkel observations were used to estimate the ratio of B to W M. callainos morphs in the general population.

120 Dominant mature adult males can be easily recognised by their behaviour and colour, but females and

121 immature males are indistinguishable and are referred to as ‘apparent females’. Hence, the number of males

122 and apparent females of each morph was counted along three 30m transects covering an area half a meter

123 each side of the line (n = 74 fish). The numbers of territory-holding males of each morph were counted in nine

124 5m2 quadrats (n = 142 fish). Although ideally the comparison should be made between non-territorial males

125 and territorial males, in practice this was not possible due to the difficulty in sexing fish without catching them.

126 However, it is likely that in the whole population, rare morph males occur at a lower frequency than rare morph

127 females (as found in other closely related species with blotch polychromatism, Lande et al. 2001; Maan & Sefc

128 2013), which would make estimates of the ratio of rare to common morph males among non-territorial fish a

129 conservative estimate; territorial W males would be present at a much lower frequency than predicted from the

130 ratio of W morph in the general population.

5 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

131 Territory distances between morphs

132 Territory maps were constructed by drawing the rocky substrate, within 5x5m string quadrats (n = 9), on dive

133 slates while snorkelling. Males frequently return to their spawning cave and focal observations allowed for

134 accurate determination of the position of this territorial focal point for each male within the quadrats. The

135 distance between the territory focal point of each male within the centre 3m2 (n = 27 B & 25 W) and closest

136 white and blue neighbour (including fish nearer the edge of the quadrat) was then measured.

137 Grazing differences between morphs

138 Each grazing action performed by focal individuals was recorded during ten minute observations of territorial

139 males and non-territorial fish (n = 9 individuals of each morph for each social status).

140

141 Field observations of aggressive interactions

142 During focal observation lasting 10 minutes per fish (n = 9 territorial males of each morph) all aggressive

143 behaviors directed towards the two conspecific morphs were recorded. The vast majority of all aggressive acts

144 recorded were ‘chases’, lateral displays were observed but rare, counts of each type of behaviour were summed

145 for analysis. While collecting data on conspecific aggression biases, aggression towards each focal fish from

146 heterospecifics was also recorded.

147

148 Laboratory aggression trials

149 To test whether there are differences in the level aggression received by blue and white morph M. callainos

150 from conspecifics and heterospecifics, three experiments were carried out using the same methods. Five

151 minute pairwise aggression trials were conducted in two replicate tanks measuring 0.9x0.3x0.3m. Each tank

152 contained a central brick refuge to act as a territory focal point, two transparent (perforated) plastic jars to hold

153 the stimulus fish, an air driven box filter and an internal heater to maintain water temperature at ca. 22-24°C.

154 Lights were kept on a 12:12 light:dark cycle. All fish were fed flake food once a day. Females and males were

155 used, but stimulus fish were always the same sex as focal fish. Focal fish were allowed at least 24h to

156 acclimatise before introduction of the stimuli and recording of focal fish behaviour began after emergence from

157 the central refuge. Individual aggressive behaviours (frontal/lateral display, quiver, lunge/butt and bite) were

158 recorded and combined to give an overall aggression count for each individual. To control for potential tank side

6 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

159 bias, two separate trials were carried out with each focal fish, each with a different stimulus fish pair and with

160 morphs swapped between sides. To avoid pseudoreplication from the re-use of focal males, before analysis an

161 average was taken of the aggressive behaviour observed in the two trials by each individual.

162 Exp. 1: Interspecific aggression biases between species

163 Firstly, conspecific aggression bias was confirmed by presenting BB M. zebra males from CB (n = 10) with pairs

164 of conspecifc BB and heterospecifc B stimulus fish.

165 Exp. 2: Intraspecific aggression biases between morphs

166 For this experiment all available M. callainos were used as focal and stimulus fish (n = 10 B male, 6 B female, 3

167 W male and 9 W female) to test for morph-specific aggression biases among conspecifics.

168 Exp. 3: Interspecifc aggression to different morphs

169 BB M. zebra focal fish from different populations (n = 12 male/ 12 female “TW”, 12 male/ 5 female “NB”, 12

170 male/ 12 female “CB”) were used to test for heterospecific aggression biases to B and W M. callainos stimulus

171 pairs. Stimulus pairs consisted of the same M. callainos used in Exp. 2.

172

173 Data analysis

174 Statistical analysis and plotting was carried out using Rstudio (v. 1.2.5033; Rstudio Team 2019) using additional

175 packages Rmisc, pscl, ggplot2, scales. General and generalized linear models were used depending on whether

176 the data originated from continuous measurements (territory distance), or counts (grazing and aggressive

177 behaviour). For the laboratory aggression experiments, trials were omitted from the analysis if the average

178 behaviour count was less than 10. In experiments where females and males were tested, sex was included in

179 the models to test whether this was a significant factor affecting aggression biases towards the stimulus

180 morphs.

181 Frequency of blue and white morphs at Thumbi West Island

182 A G-test was used to compare the actual number of territory holding W males observed with what would be

183 expected given the proportions of B and W morphs in the general population.

184 Territory distances between morphs

185 GLMs were used to test: 1) Whether there is a significant difference in the distance between focal fish and the

186 nearest neighbour of the same and different morph, and 2) differences in the average distance to B and W

7 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

187 neighbours from focal fish of the two different morphs.

188 Grazing differences between morphs

189 GLMs were used to test for effects of dominance status and morph on grazing frequency.

190 Field observations of aggressive behavior

191 Due to the small number of aggressive behaviors and relatively high number of zero counts recorded during

192 observations of aggressive encounters in the field, standard poisson regression GLMs were compared with

193 models corrected for zero-inflation. In most cases the zero inflated model was not significantly better, the

194 results reported here are therefore from the standard Poisson GLMs testing: 1) aggression directed towards

195 territorial intruders of each morph by territorial males of each morph, 2) aggression received by territorial

196 males of each morph from heterospecifics.

197 Laboratory aggression trials

198 First, GLMs were used first to test whether focal species and sex had a significant influence on the total number

199 of aggressive behaviours (overall aggressiveness of species and sexes) displayed to both stimulus fish. To control

200 for the effect of overall differences in level of aggression between species/populations/sexes/individuals, counts

201 of aggressive behaviour directed towards each of the paired stimulus fish was converted to proportion of

202 aggression. Subsequently, the following were tested: 1) whether M. zebra display a conspecific aggression bias

203 when presented with conspecific and common morph heterospecific, 2) whether among conspecifics (M.

204 callainos) overall one morph receives more aggression than the other, and whether aggression bias differs

205 among morphs and sexes, 3) whether M. zebra display an aggression bias when presented with pairs of

206 common and rare morphs of a heterospecific, and if this aggression bias differs between allopatric populations

207 of M. zebra, depending on whether they co-occur with M. callainos or not.

208

209 Results

210

211 Field observations (excluding aggression)

212 Frequency of the blue and white morphs at Thumbi West Island

213 There were significantly (G-test: G1 = 6.91, p = 0.009) more territory holding rare white (W) morph males than

214 would be expected given the proportion of W and blue (B) morph fish in the general population (fig. 2a).

8 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

215 Territory distances between morphs

216 Firstly, there is a significant difference in the average distance between same and different morph territorial

217 males (glm: t1,103 = 2.51, p = 0.0135). On average, territorial males of the same morph are found at greater

218 distance from each other (mean 1.24m) than territorial males of different morphs (mean 0.98m). While it is not

219 surprising to find that W and W are found furthest apart (mean 1.40m), as this would be the assumption based

220 on the observation of a lower frequency of W morph territorial males, if distance between morphs was only

221 based on frequency, it would also be expected that the distance between B morph males should on average be

222 the smallest distance. This is not the case: the distance between B and nearest B is on average the same (mean

223 1.07m) as, and not significantly different (glm: t1,53 = -0.008, p = 0.994) from, the distance between B and

224 nearest W. We also found that the distance between W males and their nearest B neighbour is on average the

225 smallest distance recorded between territorial males (mean 0.89m), and significantly different from the

226 distance between W territorial males and their nearest W neighbour (glm: t1,49 = 3.73, p = 0.0005) (fig. 2b).

227 Grazing differences between morphs

228 Both dominance status and morph had a significant effect on grazing frequency: Compared to territorial males,

229 non-territorial fish grazed significantly more, and regardless of social status B morph fish grazed significantly

230 less than W morph (glm; z1,35 Morph = 22.07, p <2e-16, Status = 12.18, p<2e-16) (fig. 2c).

231

232 Figure 2. Field observations at Thumbi West Island: a) There are significantly (p = 0.009) more territory holding

233 W males than would be expected given the proportion of each morph in the general population. Bars show rel-

234 ative frequency of each morph, n = 216 fish. b) Bar colours indicate neighbour colour. Differences in the dis-

235 tance of territory focal point (n = 27 B and 25 W males) among same and different morph males do not always

9 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

236 reflect the expectations based on the relative abundance of each in the population. Overall, same morph males

237 hold territories further apart from each other than different morph males (p = 0.0135). The average distance

238 between B males and the nearest B neighbour is no different from between B and nearest W neighbour (p =

239 0.994), and the distance between W territorial males and their nearest B neighbour is on average the smallest

240 distance recorded between territorial males, despite the lower frequency of W males. c) Bar colour indicate fo-

241 cal morph, n = 9 individuals of each morph and status; both territorial and non-territorial W morph fish grazed

242 significantly more than B morph fish (Morph p < 2e-16, Status p < 2e-16).

243

244 Field observations of aggressive interactions

245 Overall, B morph M. callainos territorial intruders receive significantly more aggression than their W counter-

246 parts (glm; z1,35 = -2.79, p = 0.005). However, aggression bias appears to differ between the morphs; B males

247 make significantly more attacks to other B males (glm; z1,17 = -2.93, p = 0.003), whereas W morph males show no

248 significant aggression bias (glm; z1,17 = 0.69, p = 0.493) (fig. 3a). Although there was a trend towards B fish re-

249 ceiving more aggression from heterospecifics than W fish, this difference was not significant (mean B = 1.8, W =

250 0.8, z1,17 = -1.82, p = 0.068) (fig. 3b).

251

252 Figure 3. Observation of conspecific and heterospecific aggressive interactions among territorial males in the

253 field: a) Bar colour indicates intruder colour, n = 9 focal males of each morph. B morph M. callainos territorial

254 males show significantly more aggression towards other B males than towards W males (p = 0.003), W males

255 show no significant aggression bias (p = 0.493). Overall, B morph males receive significantly more aggression (p

10 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

256 = 0.005). b) Bar colour indicates focal fish colour. There is a non-significant trend towards B morph males also

257 receiving more aggression from heterospecifics (p = 0.068).

258

259 Laboratory aggression trials

260 Differences in focal fish behaviour between species(/experiment) and sex

261 Within species, both sexes show similar levels of aggression. M. callainos overall were more aggressive (glm;

262 Focal.spec z = -10.36, p <2e-16, Sex z = 0.08, p = 0.939, fig. 4a). To test whether this difference in

263 aggressiveness was a real difference between species, or due to M. callainos being presented more often with

264 the possibility of being aggressive towards conspecific fish in these experiments, a subset of M. zebra CB and

265 M. callainos males from the three experiments was compared. M. zebra displayed a significantly higher level of

266 aggression in the experiment where the stimulus pair consisted of one conspecific and one heterospecific (Exp.

267 1) compared to the experiment where they were presented with two heterospecific stimulus fish (Exp. 3). There

268 was no significant difference, however, in the level of aggression between the species in the experiments in

269 which the stimulus pairs contained one conspecific (Exp. 1: M. zebra focal fish) or two conspecifics (Exp. 2: M.

270 callainos focal fish) stimulus fish (fig. 4b). This suggests that the presence or absence of a conspecific stimulus

271 fish contributed to the overall difference in the level of aggression observed between the species in these

272 experiments. There was therefore no evidence of species differences in intrinsic level of aggression, rather that

273 aggression among heterospecifics is lower than among conspecifics. The attack:display ratio differed between

274 sexes, but not species: Females attack more frequently and males display more (glm; Focal.spec z = 1.72, p =

275 0.089, Sex z = -4.09, p = 9.45e-05, fig. 4c).

276

11 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

278 Figure 4. Differences in levels of aggression (total aggression count) and types of aggressive behaviour (ratio of

279 attacks:displays) used among species(/experiment) and sexes. N = 29 female/ 46 male BB M. zebra (grey bars);

280 15 female/ 13 male M. callainos (blue and white bars). a) On average M. callainos were significantly more

281 aggressive (p < 2e-16), and there is no difference in the total amount of aggressive behaviour from females and

282 males (p = 0.939). b) M. zebra are only significantly less aggressive in the absence of a conspecific stimulus fish

283 (Exp. 3, p < 2e-16). N = Exp. 1, 12 M. zebra; Exp. 2, 13 M. callainos; Exp. 3 12. M. zebra. c) The attack:display

284 ratio does not differ significantly between species (p = 0.089), but males display more frequently than females

285 which use a higher proportion of attacks (p = 9.45e-05).

286

287 Aggression biases in pairwise intruder choice tests

288 As expected, M. zebra males display significantly more aggressive behaviour towards conspecifics when given

289 the choice of BB conspecific and B heterospecific stimulus males (glm; z1,19, = -16.03, p <2e-16, fig. 5, Exp 1).

290 Within M. callainos, the interaction between colour morph of focal and stimulus fish significantly affects

291 proportion of aggression received, while sex has no effect (z1,55 Interaction = 12.48, p < 2e-16, Sex = 0.00, p =

292 1.00, fig. 5, Exp 2). To further clarify whether the difference in the proportion of aggression directed to stimulus

293 fish of each colour morph is due to an own morph aggression bias (i.e. each colour morph is more aggressive to

294 other males of the same colour) or an overall common morph aggression bias (i.e. males of both colour morph

295 direct more aggression towards males of the common colour morph), stimulus type was recoded from

296 Blue/White to either Other/Own or Common/Rare: Own-morph stimulus fish received significantly greater

297 proportion of aggression than the other-morph stimulus fish, common morph stimulus fish did not receive a

298 greater proportion of aggression overall in this experiment (z1,55 Other/Own z = 12.48, p < 2e-16, Common/Rare

299 z = -0.30, p = 0.768).

300

301

12 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

302 Figure 5. In laboratory based aggression trials: Exp.1) As expected, M. zebra males bias aggression towards

303 conspecifics (n = 10, p < 2e-16). Exp. 2) Both M. callainos morphs showed a significant tendency to bias

304 aggression towards intruders of the same morph as themselves (p < 2e-16). Exp. 3) Overall the M. callainos B

305 morph stimulus fish received significantly more aggression from heterospecific M. zebra than the W morph (p <

306 2e-16).

307

308 When M. zebra were presented with the choice of M. callainos B and W stimulus pairs, overall both females

309 and males preferentially attacked the common B morph (z1,127 Stim morph = -8.79, p < 2e-16, Sex = 0.01, p =

310 0.991, fig. 5, Exp 3). However, the level of aggression (total aggression count to both morphs) and the strength

311 of aggression bias (proportion of aggression to B morph) differs between populations and sexes: M. zebra from

13 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

312 Nkhata Bay (NB) and Thumbi West (TW) were significantly more aggressive to the heterospecific stimulus fish

313 overall than those from Chiofu Bay (CB), and females overall were more aggressive than males (glm; z2,63 NB =

314 3.36, p = 0.0008, TW = 9.68, p < 2e-16, Sex = -2.62, p = 0.009, fig. 6a). In regards to strength of heterospecifc

315 common morph aggression bias, we found that NB and TW focal fish displayed a significantly lower proportion

316 of aggression towards the blue morph than those from CB (i.e. a weaker bias), and that although females were

317 more aggressive overall, they also showed a weaker common morph aggression bias than males (glm; z2,63 NB =

318 -2.50, p = 0.012, TW = -4.82, p = 1.47e-06, Sex = 2.05, p = 0.041, fig. 6b).

319

320 Figure 6. Differences in levels of aggression and aggression biases depending on M. zebra source population

321 and sex (n females/males = CB 12/12, TW 5/12, NB 12/12). a) M. zebra from CB were significantly less

322 aggressive in these experiments (compared to NB p = 0.0008, TW p < 2e-16), and females more aggressive

323 overall than males (p = 0.009). b) M. zebra from CB showed a significantly stronger common (B) morph

324 aggression bias than those from NB (p = 0.012) and TW (p = 1.47e-06), and females also displayed a weaker

325 aggression bias than males (p = 0.041).

326

327 Discussion

328 Our field observations of Maylandia callainos, a polychromatic mbuna cichlid from Lake Malawi, indicated that

329 common (blue) morph territorial intruders received more aggression than rare (white) morph intruders.

330 Pairwise intruder choice tests in a controlled laboratory setting demonstrated that males and females of each

331 morph bias aggression towards their own morph. These results add to the evidence that inherent own-morph

332 aggression biases, which result in negative frequency dependent selection on rare colour morphs, can

14 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

333 contribute to balancing selection and thereby promote the maintenance of polychromatism (Dijkstra et al.

334 2008; Lehtonen 2014; Scali et al. 2020).

335 While this and previous studies (e.g. Dijkstra et al. 2008; Dijkstra et al. 2009a; Seehausen & Schluter 2004;

336 Lehtonen 2014; Scali et al. 2020) have shown that aggression biases can be involved in stabilising

337 polychromatism, to our knowledge this is the first study to identify benefits to rare morph cichlids which may

338 result from receiving less aggression in the natural environment. We found there to be significantly more

339 territory holding rare morph males than would be expected given the proportions of the two colours in the

340 general population. Our observations suggest that the over-representation of rare morph territory owners may

341 be influenced by two factors. Firstly, blue and white males appear to have higher tolerance of each other as

342 neighbours, being found on average significantly closer to each other than blue morph individuals. Secondly,

343 both territorial and non-territorial white morph individuals spend more time feeding, which suggests that the

344 rare colour morph may benefit from lack of recognition during competition for non-mating related resources

345 (Dijkstra et al. 2008; Lehtonen 2014; Pérez i de Lanuza G et al. 2017; Scali et al. 2020).

346 Further to showing that rare morph individuals can benefit from reduced intraspecific aggression, we found

347 that a closely related ecologically similar heterospecific (Maylandia zebra) also biases aggression towards the

348 M. callainos blue morph. While these results are in conflict with those from another cichlid fish system, which

349 suggest that rare morphs may be disadvantaged by greater heterospecific aggression (Lehtonen et al. 2015),

350 given that aggression among heterospecifics is often higher among more similar coloured than more differently

351 coloured species (Genner et al. 1999; Pauers et al. 2008; Anderson & Grether 2010; Losin et al. 2016), it is not

352 surprising to find that in some cases rare morph individuals may receives less aggression from a heterospecific

353 which is more similar in colour to the common morph. We also found, however, that although heterospecific

354 females were more aggressive overall, they also showed a weaker blue morph aggression bias than males. We

355 speculate that a lower level of discrimination among morphs by heterospecific females, and the greater use of

356 direct attacks compared to display behaviours (this study and Arnott & Elwood 2009), may be due to the

357 difference in competition among females and males (i.e. greater heterospecific competition among females for

358 access to shelters among the rocks during incubation of offspring).

359 In cichlids and other taxa, laboratory studies have shown that in species which differ in colour among

360 allopatric populations, males tend to bias aggression towards males from their own population (Tyers & Turner

15 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

361 2013; Bolnick et al. 2016; Cooke & Turner 2018; Yang et al. 2018). In this study, we found that heterospecific

362 aggression also varies depending on whether a pair of species occurs in sympatry or allopatry. The level of

363 aggression (total aggression count to both M. callainos morphs) differs between M. zebra populations: M. zebra

364 from Nkhata Bay (NB) and Thumbi West (TW), which co-occur with M. callainos, were significantly more

365 aggressive to M. callainos than those from Chiofu Bay (CB), which are naïve to M. callainos. These findings

366 support the hypothesis that aggression among heterospecifics may often not simply be due to misdirected

367 aggression among species (Peiman & Robinson 2010), which would be indicated by higher levels of aggression

368 from the allopatric M. zebra population (CB). The persistence of heterospecific aggression at NB support the

369 idea that it has an adaptive function in long-term co-existing multi-species communities (Peiman & Robinson

370 2010; Losin et al. 2016). Although there are no M. callainos at Chiofu Bay, this location is home to another

371 closely-related species (M. esterae) which has blue males, and brown, orange and orange blotch females. M.

372 zebra at Chiofu Bay therefore do co-occur with a similar blue morph fish, but no white morph fish and we found

373 that the M. zebra from this location has a stronger blue-morph aggression bias than the other M. zebra

374 populations which co-occur with blue and white M. callainos. A previous study of a polymorphic frog species

375 found stronger aggression biases among morphs when they occur in allopatry compared to when they are

376 found in sympatry (Yang et al. 2018). Our results show a similar pattern in heterospecific aggression; a weaker

377 blue morph aggression bias in M. zebra populations which coexist with both colour morphs.

378 Our results indicate that a rare colour morph may benefit from lack of recognition as a resource competitor,

379 by both conspecifics and heterospecifics. This results in rare morph individuals receiving less aggression and

380 gaining improved access to territories and food. This can benefit rare morph individuals while they are rare, but

381 then what prevents them from increasing in frequency until fixation? Firstly, we found that rare (white) morph

382 individuals were more aggressive towards their own morph than they were to the common (blue) morph, which

383 would result in white morph individuals experiencing increasing levels of aggression as they became more

384 common. Secondly, the lower level of heterospecific aggression bias towards the common morph, in

385 populations with blue and white morphs, suggests that heterospecifics learn or evolve the ability to recognise

386 rare morph individuals as competitors. The ability to recognise rare morph individuals may increase as they

387 become more common: TW has the highest frequency of white morph individuals and the weakest common

388 morph aggression bias by heterospecifics. Finally, female preference for common-morph males may result in a

16 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

389 disadvantage to rare morph males (Roberts et al. 2009). The genes responsible for the expression of the

390 melanin-disrupted (“blotched”) morphs are almost always closely linked to a dominant female determiner, and

391 so are generally much more common in females. This suggests that these colour phenotypes are

392 disadvantageous to males, although they may be advantageous to females by providing increased crypsis or

393 reduced aggression from conspecifics and/or heterospecifics.

394

395 Conclusions

396 Our results support previous studies indicating that negative frequency-dependent antagonism can be

397 generated by own-morph aggression biases among conspecifics in cichlids which display polychromatism. We

398 find that heterospecifics show reduced aggression to rare morph individuals, suggesting that heterospecific

399 aggression may also facilitate invasion of rare colour morphs into a population. We identify potential

400 advantages to rare morph individuals in the field, in terms of territory and foraging.

401

402 Acknowledgements

403 This work was supported by Bangor University PhD scholarships to AMT and GMC. We thank Aaron Scott,

404 Winnie Courtene-Jones, Beth Powell for help with field recording, and James Whitman for collecting live fish.

405 We are grateful to the Department for National Parks and Wildlife in Malawi for allowing us to collect fish from

406 Lake Malawi National Park. All behavioural experiments were approved by Bangor University ethical review.

407 Thanks also to P.D.Dijkstra for and another anonymous reviewer for comments on an earlier version of this

408 manuscript.

409

410 References 411 Arnott G & Elwood RW. 2009. Gender differences in aggressive behaviour in convict cichlids. Behaviour,

412 78: 1221-1227.

413 Anderson CN & Grether GF. 2010. Interspecific aggression and character displacement of competitor

414 recognition in Hetaerina damselflies. Proceedings of the Royal Society of London B. 277: 549-555.

415 Bolnick DI, Hendrix K, Jordan LA, Veen T & Brock D. 2016. Intruder colour and light environment jointly

416 determine how nesting male stickleback respond to simulated territorial intrusions. Biology Letters. 12:

17 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

417 20160467

418 Cooke GM & Turner GF. 2018. Team colours matter when playing away from home: aggression biases in

419 geographically isolated mbuna cichlid populations. Hydrobiologia. 809: 31-40.

420 Couldridge VCK & Alexander GJ. 2002. Colour patterns and species recognition in four closely related species of

421 Lake Malawi cichlid. Behavioral Ecology. 13: 59-64.

422 Dijkstra PD, Hemelrijk CK, Seehausen O & Groothuis TGG. 2009a. Colour polymorphism and intrasexual

423 selection in assemblages of cichlid fish. Behavioural Ecology. 20: 138-144.

424 Dijkstra PD, Seehausen O & Groothuis TGG. 2008. Intrasexual competition among females and the stabilization

425 of a conspicuous colour polymorphism in a Lake Victoria cichlid fish. Proceedings of the Royal Society of

426 London B. 275: 519-526.

427 Dijkstra PD, van Dijk S, Groothuis TGG, Pierotti MER & Seehausen O. 2009b. Behavioral dominance between

428 female color morphs of a Lake Victoria cichlid fish. Behavioral Ecology. 20: 593-600.

429 Dijkstra PD, Seehausen O, Pierotti MER & Groothuis TGG. 2007. Male-male competition and speciation:

430 Aggression bias towards differently coloured rivals varies between stages of speciation in a Lake Victoria

431 cichlid species complex. Journal of Evolutionary Biology. 20: 496-502.

432 Drury JP, Cowen MC & Grether GF. 2020. Competition and hybridization drive interspecific territoriality in birds.

433 Proceedings of the National Academy of Sciences. 117: 12923-12930.

434 Genner MJ, Turner GF & Hawkins SJ. 1999. Resource control by territorial male cichlids in Lake Malawi. Journal

435 of Animal Ecology. 68: 522-529.

436 Grether GF, Losin N, Anderson CN & Okamoto K. 2009. The role of interspecific interference competition in

437 character displacement and the evolution of competitor recognition. Biological Reviews. 84: 617-635.

438 Henze MJ, Lind O, Mappes J, Rojas B & Kelber A. 2018. An aposematic colour‐polymorphic moth seen through

439 the eyes of conspecifics and predators – Sensitivity and colour discrimination in a tiger moth. Functional

440 Ecology. 32: 1797-1890.

441 Hughes KA, Houde AE, Price AC, Rodd FH. 2013. Mating advantage for rare males in wild guppy populations.

442 Nature. 503: 108-110.

443 Huxley J. 1955. Morphism and evolution. Heredity. 9: 1-52.

444 Konings A. 2007. Malawi cichlids in their natural habitat 4th edition. El Paso: Cichlid Press.

18 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

445 Jordan RC. 2008. Color-based association among heterospecifics in Lake Malawi rock-dwelling cichlids. Ethology.

446 114: 272-278.

447 Kim KW, Jackson BC, Zhang H, Toews DPL, Taylor SA, Greig EI, Lovette IJ, Liu MM, Davison A, Griffith SC, Zeng K &

448 Burke T. 2019. Genetics and evidence for balancing selection of a sex-linked colour polymorphism in a

449 songbird. Nature Communications. 10: 1852-1861.

450 Lande R, Seehausen O & van Alphen JJM. 2001. Mechanisms of rapid sympatric speciation by sex reversal and

451 sexual selection in cichlid fish. Genetica. 112-113: 435-443.

452 Lehtonen TK. 2014. Colour biases in territorial aggression in a Neotropical cichlid. Oecologia. 175: 85-93.

453 Lehtonen TK, Sowersby W & Wong BBM. 2015. Heterospecific aggression bias towards a rare colour morph.

454 Proceedings of the Royal Society of London B. 282: 20151551.

455 Losin N, Drury JP, Peiman KS, Storch C & Grether GF. 2016. The ecological and evolutionary stability of

456 interspecific territoriality. Ecology Letters. 19: 260-267.

457 Maan ME, Eshuis B, Haesler MP, Schneider MV, van Alphen JJM & Seehausen O. 2008. Color polymorphism and

458 predation in a Lake Victoria cichlid fish. Copeia. 3: 621-629.

459 Maan ME & Sefc KM. 2013. Colour variation in cichlid fish: Developmental mechanisms, selective pressures and

460 evolutionary consequences. Seminars in Cell & Developmental Biology. 24: 526-528.

461 Pauers MJ, Kapfer JM, Fendos CE & Berg CS. 2008. Aggressive biases towards similarly coloured males in Lake

462 Malawi cichlid fishes. Biology Letters. 4: 156-159.

463 Peiman KS & Robinson BW. 2010. Ecology and evolution of resource-related heterospecific aggression. The

464 Quarterly Review of Biology. 85: 133-158.

465 Pérez i de Lanuza G, Carretero MA & Font E. 2017. Intensity of male‐male competition predicts morph diversity

466 in a color polymorphic lizard. Evolution. 71: 1832-1840.

467 Ribbink AJ, Marsh BA, Marsh AC, Ribbink AC & Sharp BJ. 1983. A preliminary survey of the cichlid fishes of rocky

468 habitats in Lake Malawi. South African Journal of Zoology. 18: 149-308.

469 Roberts RB, Ser JR, Kocher TD. 2009. Sexual conflict resolved by invasion of a novel sex determiner in Lake

470 Malawi cichlid fishes. Science. 326: 998-1001.

471 RStudio Team. 2019. RStudio: Integrated Development for R. Rstudio. PBC. Boston, MA.

472 (http://www.rstudio.com/)

19 bioRxiv preprint doi: https://doi.org/10.1101/2021.04.08.439056; this version posted April 9, 2021. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC 4.0 International license.

473 Scali S, Mangiacotti M, Sacchi R, Coladonato AJ, Falaschi M, Saviano L, Rampoldi MG, Crozi M, Perotti C, Zucca F,

474 Gozzo E & Zuffi MAL. 2020. Close encounters of the three morphs: does color affect aggression in a

475 polymorphic lizard?. bioRxiv. https://doi.org/10.1101/2020.05.02.074146

476 Seehausen O & Schluter D. 2004. Male-male competition and nuptial-colour displacement as a diversifying

477 force in Lake Victoria cichlid fishes. Proceedings of the Royal Society of London B. 271: 1345-1353.

478 Seehausen O & van Alphen JJM. 1998. The effect of male coloration on female mate choice in closely related

479 Lake Victoria cichlids ( nyererei complex). Behavioral Ecology & Sociobiology. 42: 1-8.

480 Seehausen O, van Alphen JJM & Lande R. 1999. Color polymorphism and sex ratio distortion in a cichlid fish as

481 an incipient stage in sympatric speciation by sexual selection. Ecology Letters. 2: 367-378.

482 Streelman JT, Albertson RC & Kocher TD. 2003. Genome mapping of the orange blotch colour pattern in cichlid

483 fish. Molecular Ecology. 12: 2465-2471.

484 Surmacki A, Ożarowska-Nowicka A & Rosin ZM. 2013. Color polymorphism in a land snail Cepaea nemoralis

485 (Pulmonata: Helicidae) as viewed by potential avian predators. Naturwissenschaften. 100: 533-540.

486 Svensson EI. 2017. Back to basics: using colour polymorphisms to study evolutionary processes. Molecular

487 Ecology. 26: 2204-2211.

488 Takahashi Y, Yoshimura J, Morita S & Watanabe M. 2010. Negative frequency-dependent selection in female

489 colour polymorphism of a damselfly. Evolution. 64: 3620-3628.

490 Tyers AM & Turner GF. 2013. Signal and preference divergence among populations of the non-endemic basal

491 Lake Malawi cichlid fish Astatotilapia calliptera (Perciformes: Cichlidae). Biological Journal of the Linnean

492 Society. 110: 180-188.

493 Wellenreuther M, Svensson EI & Hansson B. 2014. Sexual selection and genetic colour polymorphisms in

494 . Molecular Ecology. 23: 5398-5414.

495 Yang Y, Dugas MB, Sudekum HJ, Murphy SN & Richards-Zawacki Cl. 2018. Male–male aggression is unlikely to

496 stabilize a poison frog polymorphism. Journal of Evolutionary Biology. 31: 457-468.

20